High Frequency Compression Driver Evaluation

Art Welter

Senior
Jan 11, 2011
905
45
28
Florida
My thanks to Bennett Prescott, U.S. Sales Manager for B&C speakers for guiding me to the proper driver for my application from the huge variety that B&C makes.

Thanks to Charlie Tappa of Pro Sound Service Inc. (Pro Sound Service AV Sales Service Install 877-776-7631) for providing that B&C driver for evaluation, with prior knowledge of the abuse I would subject it to.

Thanks to Jack Arnott of Assistance Audio (Welcome to Assistance Audio) for providing the BMS drivers with prior knowledge that they most likely would not fit my needs.
For years I had wanted to hear BMS drivers, which use polyester diaphragms, a material primarily associated with awful clothing from the 1970s.

Finally, thanks to Bonnie T, love of my life who put up with my obsessions over the last month in particular, and the last four years in general.

This HF compression driver evaluation was undertaken for three basic reasons:

1. Advances in materials and design has led to higher power handling which may correspond to additional available output and lower distortion.

2. The DH1AMT / DH1A drivers presently in my speaker cabinets weigh 23 pounds, a weight savings of 10 to 20 pounds per 50 pound cabinet could be realized with other drivers.

3. Most of my sound work presently is in the dry high desert of New Mexico, HF air losses in hot summer temperatures require far more acoustic power to overcome those losses compared to similar heat at high humidity.
HF air losses are in addition to the inverse distance 6 dB per doubling of distance.
For example, at 90 degrees, 80 % relative humidity, HF air loss is only 2.4 dB per 100 feet, while at 20% humidity, 22.9 dB loss occurs.
To achieve the same HF response at 100 feet in the desert would requires a 20 .5 dB increase in SPL, over 20 times the power (without considering power compression).
Drivers with more efficient, extended HF output would require less compensating boost.

In the past two decades, LF driver technology has made large advances in linear excursion capability and power handling, resulting in more clean LF output per driver, although requiring more power to achieve that goal.

Thiele -Small parameters for cone transducers make comparison of cone speakers relatively easy, allow allowing reasonably reliable predictions of response in a particular cabinet alignment.

HF compression driver manufacturers do not publish TS parameters, at best one gets power, Fs (resonant frequency) diaphragm size and composition, exit diameter and response on a specific horn or plane wave tube, sometimes with harmonic distortion figures derived from a single frequency sweep at a small fraction of rated power.

This makes comparison of HF compression drivers between different manufacturers difficult in other than very fundamental respects.
Even comparing within the same company can be difficult, the newer EV DH7 can’t be compared to the older DH1AMT on paper, they used different horns in testing.

Having heard dozens of different HF drivers on dozens of different horns in different venues ranging from home theaters to arenas, with different arrays driven with various playback material and mixes by different engineers mixing different bands through different consoles, processing, and amplification, about the only conclusion drawn were various HF drivers can sound quite good or lousy.

The EV DH1AMT and DH1A HF drivers presently in my PA were chosen back in 1992 after extensive side by side testing comparing them to the JBL 2445 and 2425 drivers in use at the time.
A single DH1A had more extended, clean response than the pair of JBLs, at a considerable weight, size, and cost advantage, both in reduction of driver cost and elimination of a crossover point and amp channels.

In 2000, after comparison, Jack Arnott replaced his JBL 2450 drivers with BMS, and later became the North American distributor for BMS product.

I designed a series of tests that involves elimination of as many variables described above to determine how a variety of drivers compare to each other at different drive levels, all driven with the same program material through the same electronics chain. Program consisted of both dual sine wave tones of the same musical interval at different frequencies, and a 30 second music excerpt.
The dual sine wave tones have accompanying RTA screen shots so the difference in each driver’s harmonic and intermodulation distortion can be compared visually as well.

The results of the tests were recorded and normalized for level and are available for download.

The “Full Monty” report with all the facts and figures is here:
High Frequency Compression Driver Evaluation - diyAudio

Since you are a busy professional, I’ll cut to the chase and give you the results of what consumed 50.25 hours of testing, then another 84.75 hours to mix, analyze, write up and post the findings, a total of 135 hours of work in this project spent over the last month.

To be continued next post.

1A = EVDH1A
02= Eminence PSD2002
50= BMS 4550
52= BMS 4552
82+ B&C DE82
“DRPA630HiOutOnly” is a recording of the DSP HF output driven by a Phillips CD player.
The recordings below are an excerpt from Ed Kabotie’s song, “7 Cities of Gold”
Ed Kabotie Freedom Songs - DreamCatcher.com
with voice, acoustic guitar and wood flute used to compare the output of the various HF drivers all fitted to the same type horn.

The drivers were equalized as flat as possible from 630 to 16K using the three PEQ and 1/3 octave filters afforded by the DBX DriveRack PA. Each driver also used a specific delay compensation to align it with the LF track, slightly different delay and EQ were required for the 1250 Hz alignment.
The process was akin to aligning a dozen different PA systems, no small feat in itself.
The process “babied” the 1” drivers, allowing for the LF track to carry some extra “weight” around the 630 Hz crossover point. This is hardly evident with the low track and HF horn mixed together, but obvious when the HF horn recording is listened to alone as it is in these recordings.

The actual SPL and power levels for the recordings are:

1An0630 0, 630 Hz (7.3 watts, 116.7 dBA peak)
1Ap1363 +13, 630 Hz ( 146 watts, 128.9 dBA peak)
1Ap1763 +17, 630 Hz (367 watts, 132.4 dBA peak, oops)
1Ap1312 +13, 1250 Hz (73 watts, 127.8 dBA peak)

02n0630 0, 630 Hz (7.3 watts, 110.3 dBA peak)
02p0125 0, 1250 Hz ( 11.5 watts 113.8 dBA peak)
02p10125 +10, 1250 Hz (115 watts 118.9 dBA peak)
02p15 125 +15, 1250 Hz (367 watts 125.8 dBA peak, oops)

50n0630 0, 630 Hz (7.3 watts, 114 dBA peak)
50p13125 +13, dBA 1250 Hz (36.6 watts, 125.5 dBA peak)

52n0630 0, 630 Hz (7.3 watts, 116.3 dBA peak)
52p13125 +13, 1250 Hz( 36.6 watts 124.6 dBA peak)

82n0630 0, dBA 630 Hz (7.3 watt 113.2 dBA peak)
82p1363 +13, 630 Hz (146 watts 126.3 dBA peak)
82p1312 +13, 1250 Hz (73 watts 121.6 peak)

Not wanting to run the risk of burning up this antique driver, it was given less power:
PAp363 +3, dBA 630 Hz ( 14.6 watts 96.3 dBA peak)
PAp912 +9, dBA 1250 Hz (46 watts, 99.5 dBA peak)

More recordings are included in Posts # 2 and 3.
I suggest downloading "DRPA630HiOutOnly" (the high output of the crossover used for testing the HF drivers), first as a comparison to hear what the high end of the song sounds like before being recorded from the output of the HF drivers.
The sound files are in mp3 format, to open them change the suffix < .pdf > to <.mp3 >, then open with your mp3 app.

Art Welter
 

Attachments

  • HF 200-2K.png
    HF 200-2K.png
    92 KB · Views: 10
  • 02n0630HiOnly.pdf
    549.9 KB · Views: 6
  • 02p0125HiOnly.pdf
    554.5 KB · Views: 5
  • 50n0630HiOnly.pdf
    566.8 KB · Views: 2
Last edited:
Re: High Frequency Compression Driver Evaluation

The Facts:

All the drivers tested are nominally rated as 16 ohm. “DCR” was the DC resistance measured (in ohms), for relative comparison. Sorry for the run on mess, tabs (and more than single spaces in a line) vanish after editing.

Driver name, Test name, AES, Peak, Rated Range, Exit, Diaphragm, DCR, Angle, Lb.
BMS 4550, 50, 80, 450, 500-20 kHz, 1”, 1.75”, Polyester, 12.3, 12, 4.95
BMS 4552, 52, 80, 450, 500-20 kHz, 1”, 1.75”, Polyester, 12.3, 12, 1.83
BCDE82, 82, 110, 220, 500-18 kHz, 1.4”, 3”, Tit.,Polyimide 8, <1, 9.9
PSD2002, 02, 80, N/A, 1.2K-20 kHz, 1”, 2”, Titanium, 11.0, 10, 5.0
EVDH1A, 1A, 50, 200, 500-20 kHz, 1.4-2”, 3”, Titanium, 10.4, 14, 23.5
Jensen D-30, PA, 30?, ?, ?, .75”, 2.25”?, Aluminum, 11.40, <0, 2.5

The tests used Welter Systems Maltese conical expansion horns (13 x 13 degree horn angle) with adapters to fit the various compression driver exits. The narrow horn results in an on axis sensitivity around 6 dB more than most typical 90 degree horns, the 2 meter results would be similar to one meter results on some horns.
Care was taken to insure a smooth transition between driver, adapter and horns, requiring a fair amount of drilling, filing and cursing those that made poorly aligned adapters and drivers that have unique bolt patterns.
The Maltese horn exhibits increasing “ripple” (frequency response dips and peaks) below 1000 Hz, the ripple reduces to less than .25 dB by 2000 Hz, where the frequency response differences between drivers become large, in some cases over 20 dB. The low frequency ripple is less than the unit to unit driver frequency response variance.

1) Size does matter regarding LF output, there is no replacement for displacement. The larger diaphragm drivers have more output below 1250 Hz, the 3” diaphragm driver’s response extending to 200 Hz. The larger diaphragms provide more output with less distortion.
The smallest driver, the BMS 4552 had the least LF, followed by the 4550, the Jensen Hypex D-30, and the Eminence PSD2002.
The EV DH1A and the B&C DE82 were neck and neck in LF response, the DE82 having slightly more output below 1000 Hz, other than the DH1A having slightly more output below 250 Hz.

The BMS drivers are only capable of low distortion output with a 630 Hz crossover frequency at a level lower than one watt. BMS makes coaxial 2” exit drivers with better LF response.
The BMS also had more IM distortion than the other drivers, around 14%.
The old Jensen half the distortion, 16%.
The inexpensive PSD2002 was a surprise, with 6 dB more power, it hit 95.7 dB at 530 Hz with 8% 2ndHD.
The B&C DE82 at +6 dB did 98.9 dB, 6% 2ndHD.
At +9 dB, 101.9 dB 7% 2ndHD, the ultra-heavyweight EV DH1A had a clear 3 dB “win”.

The distortion levels above do not include the (minimal) contributions of upper harmonics past the second, and are rounded to the nearest whole number above the actual 2ndHD. Anyone willing to crunch the numbers further is welcome to look at the RTA charts and derive finer resolution. Having done so in the past to prove the point that 3rd, 4th etc. harmonics, unless they exceed the second, change THD very little.
Odd order harmonics tend to sound worse (at least from a western musical standpoint) and there are differences in their levels between the drivers.

Harmonics of the “Mid” test, using 1046 & 1865 Hz sine wave tones occurred at 819 Hz ( IM sub harmonic component) and around 2092, 2660, 2920, 3138, 3450, 3730, 3980, 4730, 5010, 5620, 5790, 7760, 8410 Hz.
The Jensen Hypex PA30 driver, lacking a phase plug, had an additional 16 odd harmonics occur.

The BMS drivers again had the highest IM content. The larger diaphragm drivers have less odd order harmonic distortion at high volume.

Harmonics of the ”Hi” test, using 2093 & 3729 Hz tones sine wave tones occurred at 1636 Hz (IM) and around 4190, 5360, 5822, 6280, 7460, 7915, 9100, 9550, 10008, 11190, 11640, 12100, 13280, 13740, 14920, 15380, 16290, 17010 Hz.
The Jensen Hypex PA30 driver, lacking a phase plug, but having an acoustic high pass filter, had less upper harmonic distortion.
The Eminence PSD2002, which has little output above 12500 Hz, hit a slightly higher SPL than the other 1” drivers with similar distortion level.

2) The dual sine wave “mid” test using 1046 &1865 Hz came out with a different order at the bottom in a close race, the DH1a again the top driver in terms of clean output capability.
02dsMid+3 104 dB, 7% 2ndHD
PAdsMid+6 104 dB, 7% 2ndHD
52dsMid+3 104.8 dB, 10 % 2ndHD
50dsMid+3 105.7 dB, 8 % 2ndHD
82dsMid+3 106.6 dB, 6% 2ndHD
1AdsMid+9 112.1 dB, 10% 2ndHD

3) The “Hi” test using 2093 & 3729 Hz resulted in less disparity between the drivers, the EV still had the most output:
PAdsHi 106.4 dB <4 % 2ndHD
02dsHi+6 108.4 dB, 4% 2ndHD
52dsHi 110.3 dB, < 6% 2ndHD
50dsHi 110.7 dB, 6% 2ndHD
82dsHi+6 111.1 dB, 5% 2ndHD
1AdsHi+3 112.9 dB, < 6% 2ndHD

4) The B&C DE82 had the flattest response, +/- 3 dB from 644 to 5000 hz, with a peaked rising response above. The BMS and EV drivers have more output in the 1600 to 8000 Hz range.
The Eminence PSD2002 has a rough response, and little output above 12.5 kHz.
The Jensen Hypex rolls off at 2500 Hz, with a shelved upper response, more explanation in it’s test section below.

5) Compression drivers “compress” the output of a relatively large diaphragm in to a much smaller throat volume, a 3” diaphragm with a 1.4” exit at minimum has a 4.55/1 compression ratio, ratios as high as 10/1 are common. This compression results in high throat SPL, at the upper power levels the drivers were tested at the non- linearity of air is responsible for a significant (though undetermined percentage) of the of distortion measured, that non- linear distortion is unavoidable in any compression driver design.
With approximately one watt of input, the B&C DE82 measures 110 dBA at 2 meters, that level results in 143.7 dBA at the bug screen, still one inch from the throat. Using a conservative value of 20 dB above that level (well within the AES power rating) throat SPL will reach a level of 168 dB SPL.
I have not measured any closer than the bug screen, one other fact found during testing is pebbles and dust contain enough ferrous material to stick to the driver’s magnets, requiring cleaning with sticky tape to remove. Removal of the bug screen would be an invitation to introduce magnetically attracted “trash” in the gap.
Examples of air non-linearity introduced by the mixing of two ultrasonic tones at 104 dB SPL are available in posts #179 #182 of:
What are benefits of adding HF driver 7khz up? - Page 18 - diyAudio

Opinions:
Considering the disparity between size, weight, diaphragm material, and design, the constraints and compromises inherent in each approach, once equalized for flat on-axis response there was surprisingly little difference in the end results of listening tests.
The dual sine wave tests made hearing the differences between the drivers easiest, simply stated, the EV DH1A has more output available with less audible distortion than the other drivers. The B&C DE82, also using a 3” diaphragm, was the next choice in terms of output vs. distortion.

The 3” diaphragm drivers had large advantage in terms of low frequency output, they also had close to a 3 dB advantage in terms of HF output.
Though the BMS drivers were more sensitive in high frequency output, the smaller diaphragms and throats “run out of gas” about 3 dB SPL before the larger diaphragms.
The Eminence PSD 2002 wins the low end 1” competition, the BMS the mid -high.
I expect opinions to vary widely on this, the low end distortion in the BMS may be construed as “warmth”, and although the PSD2002, has more undistorted LF, when pushed hard sounds more “grindy” compared to the BMS.
The Jensen Hypex, once equalized, sounded surprisingly good considering it’s abysmal raw response.

With a music playback, when pushed hard, even with a 1250 crossover, the "s" sounds in the word “cities” seemed to get the most “spitty” with the BMS drivers, and the harmonics of the flute seemed a bit odd. The BMS drivers have the smallest diaphragms of the drivers tested, whether this type of distortion is simply the result of the higher throat SPL required by the smaller diaphragm and exit diameter, or the driver design itself is an unanswered question.

The upward masking from lower frequencies crossover masks those distortion effects to a large degree.

While recording I only listened to the HF output, what I heard often reminded me of distorted screams heard in “B” grade horror movies, I was surprised to find how much distortion can be tolerated in the presence of the low frequency output.

Conclusions:
All of the drivers tested showed many desirable characteristics and some defects.

The Jensen Hypex is an antique, no longer available except on the used market. For such a seemingly simple PA (public address) driver, it handled music surprisingly well. Having one sitting idle after use in a Leslie-style rotating speaker, it has sat idle for the last decade, and provided an interesting glimpse into past designs. I had assumed from it’s rolled off upper response it used a phenolic diaphragm as most thread on PA drivers, finding it used an aluminum diaphragm was a surprise.

The Eminence PSD2002, the cheapest driver of the bunch, had far more undistorted lower frequency output than the other one inch exit drivers, though upper response was lacking, and more “heroic” equalization was required throughout it’s operating range. The PSD2002 is used as an OEM driver by many manufacturers, it provides fairly bulletproof operation. Having owned many of these drivers, have found unit to unit consistency to be quite poor, the driver tested was the best of four units I still own.

The BMS 1” drivers have extended clean HF response, but require fairly high crossover points if high volume, low distortion response is desired. The BMS 4552 is by far the highest output for size and weight of the drivers tested. The small size lends itself to line array applications, the close driver spacing reducing the complexity of the driver/wave guide interface. In quantity, it is also a great choice for combating HF air attenuation, sixty four units have been coupled to single horns using Tom Danley’s layered combiner. They have a “classic” HF driver response that would work well with many “old school” horns, yet have upper HF response extending above human audibility.

The B&C DE82 has most all the elements I look for in a compression driver: low distortion, extended bandwidth, high output capability, relatively low weight and cost. It was not the most efficient driver, but was the flattest overall, and appears to handle the extra power it requires over the others with no problems. The only small complaint are some response glitches at the very top end of it’s response, BMS makes other drivers that (other than price) address those small HF issues.

I was hoping to find a lighter weight replacement with sound quality exceeding the Electro-Voice DH1A, but in this series of tests it simply was the best sounding driver in every metric I can think of: low distortion, extended bandwidth, high output capability, and durability with prolonged unit to unit consistency.

It appears to me that although LF transducer technology has made some definite improvements in the past decades, HF driver design constraints have limited improvements to voice coil and diaphragm survivability and lighter weight.

Of course, if any manufacturer’s would like to send me pairs of free compression driver production samples to evaluate, I’ll be happy to re-evaluate my last conclusion upon finding a driver with better overall performance than the EV DH1A or DH1AMT.
Until then the ten DH1AMT and single DH1A drivers I own will remain in their present cabinets.


Since as a professional in the audio feild you probably care far more about the performance of a driver when pushed hard, several examples will be posted.

The musical recordings include the output of the high frequency horn only, a full range mix of the same files are available, recorded at the same and lower levels down to just under a watt:
High Frequency Compression Driver Evaluation - diyAudio

1A = EVDH1A
02= Eminence PSD2002
50= BMS 4550
52= BMS 4552
82= B&C DE82
“DRPA630HiOutOnly” is a recording of the DSP HF output driven by a Phillips CD player.
The recordings below are an excerpt from Ed Kabotie’s song, “7 Cities of Gold”
Ed Kabotie Freedom Songs - DreamCatcher.com
with voice, acoustic guitar and wood flute used to compare the output of the various HF drivers all fitted to the same type horn.

The drivers were equalized as flat as possible from 630 to 16K using the three PEQ and 1/3 octave filters afforded by the DBX DriveRack PA. Each driver also used a specific delay compensation to align it with the LF track, slightly different delay and EQ were required for the 1250 Hz alignment.
The process was akin to aligning a dozen different PA systems, no small feat in itself.
The process “babied” the 1” drivers, allowing for the LF track to carry some extra “weight” around the 630 Hz crossover point. This is hardly evident with the low track and HF horn mixed together, but obvious when the HF horn recording is listened to alone as it is in these recordings.

The actual SPL and power levels for the recordings are:

1An0630 0, 630 Hz (7.3 watts, 116.7 dBA peak)
1Ap1363 +13, 630 Hz ( 146 watts, 128.9 dBA peak)
1Ap1763 +17, 630 Hz (367 watts, 132.4 dBA peak, oops)
1Ap1312 +13, 1250 Hz (73 watts, 127.8 dBA peak)

02n0630 0, 630 Hz (7.3 watts, 110.3 dBA peak)
02p0125 0, 1250 Hz ( 11.5 watts 113.8 dBA peak)
02p10125 +10, 1250 Hz (115 watts 118.9 dBA peak)
02p15 125 +15, 1250 Hz (367 watts 125.8 dBA peak, oops)

50n0630 0, 630 Hz (7.3 watts, 114 dBA peak)
50p13125 +13, dBA 1250 Hz (36.6 watts, 125.5 dBA peak)

52n0630 0, 630 Hz (7.3 watts, 116.3 dBA peak)
52p13125 +13, 1250 Hz( 36.6 watts 124.6 dBA peak)

82n0630 0, dBA 630 Hz (7.3 watt 113.2 dBA peak)
82p1363 +13, 630 Hz (146 watts 126.3 dBA peak)
82p1312 +13, 1250 Hz (73 watts 121.6 peak)

Not wanting to run the risk of burning up this antique driver, it was given less power:
PAp363 +3, dBA 630 Hz ( 14.6 watts 96.3 dBA peak)
PAp912 +9, dBA 1250 Hz (46 watts, 99.5 dBA peak)

The files are in mp3 format, to open them change the suffix < .pdf > to <.mp3 > they will open with your mp3 app.

Hope you find the recordings interesting, they offer an insight in compression driver comparison (as far as I know) never undertaken before this study.

Art Welter
 

Attachments

  • 52n0630HiOnly.pdf
    564.2 KB · Views: 5
  • 52p13125HiOnly.pdf
    539.7 KB · Views: 2
  • 82p1363HiOnly.pdf
    597.4 KB · Views: 1
  • DRPA630HiOutOnly.pdf
    575.9 KB · Views: 1
  • PAp363 HiOnly.pdf
    554.5 KB · Views: 2
Last edited:
Re: High Frequency Compression Driver Evaluation

More recordings, description of them is at the bottom of post #1 .

The files are in mp3 format, to open them change the suffix < .pdf > to <.mp3 > they will open with your mp3 app.


Art
 

Attachments

  • 1Ap1363HiOnly.pdf
    578 KB · Views: 1
  • 1Ap1763 HiOnly.pdf
    563.7 KB · Views: 2
  • 1APn0630HiOnly.pdf
    542.3 KB · Views: 2
  • 1APp1312HiOnly.pdf
    591.2 KB · Views: 1
  • 02p15 125HiOnly.pdf
    575.4 KB · Views: 1
  • 02p0125HiOnly.pdf
    554.5 KB · Views: 1
  • 02p10125HiOnly.pdf
    564.2 KB · Views: 1
  • 50p13125HiOnly.pdf
    577.5 KB · Views: 2
  • 52p13125HiOnly.pdf
    539.7 KB · Views: 2
  • 82n0630 HiOnly.pdf
    591.2 KB · Views: 1
  • PAp912HiOnly.pdf
    586.1 KB · Views: 1
Re: High Frequency Compression Driver Evaluation

Thank you very, very much for posting this.

One question, and this is directed more to Bennett. Why doesn't B&C publish distortion plots for their HF drivers? I notice JBL and BMS do. (Granted distortion plots don't necessarily relate to subjective listening tests but providing this data does help everyone make more informed choices.)

Cheers,
Michael
 
Re: High Frequency Compression Driver Evaluation

I was surprised to find how much distortion can be tolerated in the presence of the low frequency output.

I could not resist quoting this Art.

An extensive discussion could be had about our hearing as it relates to distortion and the "masking" that LF signals can have on its perception. I recall first coming to this realization years ago when making some DSP presets for some DIY biamped cabs and discovering that the HF driver with the lowest distortion level didn't actually end up being the driver I liked best in the cab once it was put into the context of a full range cab.
 
Last edited:
Re: High Frequency Compression Driver Evaluation

One question, and this is directed more to Bennett. Why doesn't B&C publish distortion plots for their HF drivers? I notice JBL and BMS do. (Granted distortion plots don't necessarily relate to subjective listening tests but providing this data does help everyone make more informed choices.)

Michael,

Probably to keep spec sheets concise, although I don't actually know. We don't publish a great deal of the data we take for each product, all of this is of course available for the asking.
 
Re: High Frequency Compression Driver Evaluation

Michael,

Probably to keep spec sheets concise, although I don't actually know. We don't publish a great deal of the data we take for each product, all of this is of course available for the asking.
Bennett,

Sensitivity and frequency response of drivers on different horns and AES power ratings unfortunately only give a limited amount of comparative data.

BMS includes distortion specifications on their single page spec sheet, which is more concise than JBL and EV spec sheets in general.
Comparing the EV DH1A and the BMS 4550 and 4552 distortion specs it was obvious that the DH1A was going to smoke them in the lower response, less obvious that it would also surpass their HF output in terms of low distortion at high level.
I still was interested in how they sound, Jack Arnott graciously let me test the BMS drivers with the understanding they would most likely not result in a sale.

The failure of any manufacturer to provide distortion specifications at anything near rated AES power and lack of any readily available distortion data from B&C was the reason I felt the need to conduct a driver comparison myself.
Even though the tests took an enormous amount of time, I felt they would have been worth it had I found a driver that performed better than the 20 year old EV drivers I currently use.

Could you please provide distortion specifications for whatever the lowest distortion 3" diaphragm, 1.4"/1.5"/2" exit driver B&C currently has in their lineup.

I would like to compare them to the DE82 and DE82TN, the distortion spec sheet you sent was at conducted using 8 volts, 8 watts for 8 ohm drivers rated at 80 and 110 watts AES.
The amplitude response was missing in those tests, it would be helpful.

Do you have any distortion data for the drivers at their rated AES power?

Thanks,

Art
 
Re: High Frequency Compression Driver Evaluation

I could not resist quoting this Art.

An extensive discussion could be had about our hearing as it relates to distortion and the "masking" that LF signals can have on its perception. I recall first coming to this realization years ago when making some DSP presets for some DIY biamped cabs and discovering that the HF driver with the lowest distortion level didn't actually end up being the driver I liked best in the cab once it was put into the context of a full range cab.

I'd also be very interested to follow a thread on masking in relation to drivers.

Much of my work in the last 14 years has been in audio coder research - MPEG2-AAC, AC3, E-AC3. The coders make significant use of frequency masking our ears to operate and > 10:1 compression ratios. At any given moment in time, the coder calculates a masking threshold as a function frequency, and quantizes the frequency spectrum near that threshold. Any signals lower in level than the mask are thrown away. The bulk of the frequency masking is upwards - i.e. lower frequencies mask upper frequencies. Generally speaking the degree of masking increases with SPL - that is for a given frequency, the masking shape doesn't just rise with the SPL at that frequency, the shape widens. So at higher SPL's, more of the signal will be masked in the ear.

More recently coders have pushed compression ratios even high by incorporating spectral replication. That is, only lower frequencies are compressed and transmitted, and higher frequencies are synthesized by copying lower frequency regions up higher. While this may sound extreme, these techniques are backed by a lot of double blind listening tests with expert listeners. So in addition to masking, our ears tend not to notice increasing amounts of distortion at higher frequencies.
 
Last edited:
Re: High Frequency Compression Driver Evaluation

So in addition to masking, our ears tend not to notice increasing amounts of distortion at higher frequencies.

That's interesting. One of my biggest complaints about a lot of audio codecs is how much distortion in the higher frequencies I actually do hear. Granted it has much improved over the original mp3 format. But I do hear it even with AAC at the lower to mid bit rates. I suppose many folks don't notice it but I do, namely on cymbals and high hats. That phasey-swishy sound. Nowadays I use lossless AAC when encoding my own stuff. If you buy from the iTunes store, you're stuck with 256k AAC which is adequate for a lot of stuff but not always. It depends on the musical content as far as how bad the artifacts are.
 
Re: High Frequency Compression Driver Evaluation

I'd also be very interested to follow a thread on masking in relation to drivers.

Much of my work in the last 14 years has been in audio coder research - MPEG2-AAC, AC3, E-AC3. The coders make significant use of frequency masking our ears to operate and > 10:1 compression ratios.

More recently coders have pushed compression ratios even high by incorporating spectral replication. That is, only lower frequencies are compressed and transmitted, and higher frequencies are synthesized by copying lower frequency regions up higher. While this may sound extreme, these techniques are backed by a lot of double blind listening tests with expert listeners. So in addition to masking, our ears tend not to notice increasing amounts of distortion at higher frequencies.
Interesting.

The recordings were all made using an Alesis Masterlink, which makes standard Redbook CDs.

The distortion in the MP3 "bounce" recordings seems as apparent to me as listening to the original.

When you compare the DRPA (which did not go through a horn /driver) recording to the others, especially the high drive level ones, don't you hear the difference?

Art
 
Re: High Frequency Compression Driver Evaluation

That's interesting. One of my biggest complaints about a lot of audio codecs is how much distortion in the higher frequencies I actually do hear. Granted it has much improved over the original mp3 format. But I do hear it even with AAC at the lower to mid bit rates. I suppose many folks don't notice it but I do, namely on cymbals and high hats. That phasey-swishy sound. Nowadays I use lossless AAC when encoding my own stuff. If you buy from the iTunes store, you're stuck with 256k AAC which is adequate for a lot of stuff but not always. It depends on the musical content as far as how bad the artifacts are.

This work is very old and sometimes misunderstood... masking is related to human audition, part how our ears physically work, and more how our brains interpret the data from our ears.

One misunderstood relationship is that HF harmonic distortion is inaudible since the distortion components fall at above audible frequencies. HF nonlinearity can also cause LF IMD products that we can hear.

In the case of your swishy sounding cymbals, there may be HF content (distortion or actual signal) folding back down into the audible passband (aliasing) because of time sampling (Nyquist et al) or IM distortion, while IM usually sounds like muddy LF grunge. Cymbals don't conveniently stop making output at 20kHz, so close mic'd they will generate lots of very HF energy that must be filtered before coding.

There has long been a tug of war between what consumers will tolerate for audio performance and what they will pay for. There are many experts critical of current accepted "studio quality" while consumers have rejected several past attempts to improve upon decades old CD quality.

The consumer pretty much decides for better or worse, and some of the expert's complaints seem pretty thin (IMO). While other "experts" have been pushing substandard audio on consumers for decades. I have limited sympathy for consumers when I listen to car's driving by my house with clearly distorted music playing. YMMV

JR
 
Re: High Frequency Compression Driver Evaluation

I find this entire thread interesting. Dave Gunness, as far as I'm aware, can simply build an FIR filter and cancel out any distortion.

I once asked him about the non-linearity of drivers, since even though an FIR can cancel the distortion, distortion isn't constant with output, so he'd need a dynamic FIR to properly take care of the issue across the range of output levels the driver could handle. Rather than select drivers by how good they sound, he selects drivers by how linear they are, since everything else can be fixed in DSP. Who would have thought!?

Hope I'm not giving away any trade secrets here. If I am, please remove this post.
 
Re: High Frequency Compression Driver Evaluation

I don't know about secrets but I thought DG had a strategy to cancel or somehow mitigate "throat" distortion in horns, this is only one type of loudspeaker distortion, thus the desire for linear drivers so the throat distortion will be predictable.

I am not a speaker guy... does it show?

JR
 
Re: High Frequency Compression Driver Evaluation

Hello Art,
Wow, what a lot of work you put into this. I appreciate that.

If I am ever to meet you, I feel I owe to pay for your wife's dinner.
(I hope this came out right. Just that I want to thank her for her patience with you and this project.)

Lot of thought here, and more so as the thread expands. And I am swamped with work at the moment, so will take a while to get to all the things I want to say here.

First off, proclaimer. I use BMS drivers. And disclaimer, I sell them. Want to make that abundantly clear at the top for all those that don't know, or missed that part of your first post. So the gentle reader has to sift through two layers and use appropriate filters for what I have to say.

I wanted to address the size of the drivers, in relation to V(oice) C(oil) size. You noted that the VC on the BMS was the smallest in the group, 1.75", and it indeed has the lowest surface area, and one of the reasons it has less "low end", or more realistically, does not play as low.

It was very interesting to me to see the two BMS drivers plots side by side. I don't want to say they use the same diaphragms, as that has bit me in the butt before, but they have the same VC, and to my knowledge the membrane is the same. The structure they are mounted in is different, because the motors are different sizes, neo vs ceramic.

Because the 4552 (neo) is smaller than the 4550 (ceramic) the throat opens up faster and has a bigger angle at the mouth of the driver than the 4550. So to me, this is the main difference between the drivers, and why the 4552 does not go as low as the 4550. So there are other things at play besides the size of the diaphragm.

Regarding the VC size, and its relation to the surface area of the diaphragm. It is important to note that the VC in BMS drivers is in the middle of the diaphragm, because it is a ring radiator design, not at the very outside as with a traditional compression driver. As such, more than 50% of the surface area of the diaphragm is outside the VC, and there is more surface area than a conventional driver with a 1.75" VC. Also, in the same vein, there is less distance for the VC to support, or control at any one time.

More on this later, I have to go.

Thanks again for the time, energy and expertise. And the follow up conversations.

Regards, Jack
 
Re: High Frequency Compression Driver Evaluation

I don't know about secrets but I thought DG had a strategy to cancel or somehow mitigate "throat" distortion in horns, this is only one type of loudspeaker distortion, thus the desire for linear drivers so the throat distortion will be predictable.

I am not a speaker guy... does it show?

JR

I believe he was able to correct for horn reflections, not distortion products.

http://fulcrum-acoustic.com/wordpre...ponse-with-digital-signal-processing-2005.pdf has some details
 
Re: High Frequency Compression Driver Evaluation

Hello Art,
Wow, what a lot of work you put into this. I appreciate that.

I wanted to address the size of the drivers, in relation to V(oice) C(oil) size. You noted that the VC on the BMS was the smallest in the group, 1.75", and it indeed has the lowest surface area, and one of the reasons it has less "low end", or more realistically, does not play as low.

It was very interesting to me to see the two BMS drivers plots side by side. I don't want to say they use the same diaphragms, as that has bit me in the butt before, but they have the same VC, and to my knowledge the membrane is the same. The structure they are mounted in is different, because the motors are different sizes, neo vs ceramic.

Because the 4552 (neo) is smaller than the 4550 (ceramic) the throat opens up faster and has a bigger angle at the mouth of the driver than the 4550. So to me, this is the main difference between the drivers, and why the 4552 does not go as low as the 4550. So there are other things at play besides the size of the diaphragm.

Regarding the VC size, and its relation to the surface area of the diaphragm. It is important to note that the VC in BMS drivers is in the middle of the diaphragm, because it is a ring radiator design, not at the very outside as with a traditional compression driver. As such, more than 50% of the surface area of the diaphragm is outside the VC, and there is more surface area than a conventional driver with a 1.75" VC. Also, in the same vein, there is less distance for the VC to support, or control at any one time.

Regards, Jack

Jack,

Thank you again for the loan of the interesting drivers.
I was unaware these drivers use a ring radiator type diaphragm.
Could you post a photo of the diaphragm?

The low frequency response output was considerably less (and LF distortion much higher) for the 1” exit BMS drivers than the 1” Eminence PSD 2002, I would think the diaphragm surface area (SD) must be quite a bit less.

In my rough measurement of the throat angle of the 4550 and 4552 they looked identical.

Assuming the diaphragm and voice coil are the same, I would attribute the lesser LF response to the smaller compression chamber the 4552 has compared to the 4550.

Could you please check with the factory and find the exact exit angle for both drivers, and SD of each diaphragm?

Art
 
Re: High Frequency Compression Driver Evaluation

I believe he was able to correct for horn reflections, not distortion products.

http://fulcrum-acoustic.com/wordpre...ponse-with-digital-signal-processing-2005.pdf has some details

You are correct, Gunness focusing addresses four problems, distortion is not one of the correctable problems:
1) All phase plug designs produce significant smearing of the transient response.
A significant fraction of the sound energy arriving at a phase plug opening will either continue past it or reflect back from it; in either case arriving later at other phase plug slots where the sound is divided again, ad infinitum.
A carefully constructed filter falls short of perfectly correcting the axial response, but it improves it significantly, while also significantly improving the off axis response.

2) A second loudspeaker behavior, which yields well to digital preconditioning, is horn resonance.

3) A third behavior, which also yields to digital preconditioning, is cone resonance.

4) A fourth behavior is non-linear phase response of the summed crossover.
The essence of the ilter to correct that defect is the high frequencies are delayed relative to the low frequencies, which counteracts the minimum phase crossover’s effect of delaying the low frequencies relative to the highs. This requires an added high frequency latency of .7 ms, the full length of the FIR filter. A nearly perfect approximation can be achieved with 1 ms or more.


All four combined can indeed make a speaker sound much better.

My driver comparison focused on the uncorrectable problem of distortion.

Uncorrectable, as evidenced by an older driver design developed more than two decades ago while Dave Gunness was still working for Electro- Voice having less distortion than more current designs.

The distortion inherent in HF compression drivers is a tough nut to crack.

The lack of progress in reducing distortion may be why Bennett, who is usually quite prompt, has not responded to my requests for further distortion information from B&C in post #7.

Art
 
Re: High Frequency Compression Driver Evaluation

OK, I stand corrected... throat distortion is a phenomenon related to the air going non linear.

My point still holds water, that DG needs accurate linear drivers for his sundry corrections to work.

I'll stop talking about stuff i shouldn't be for now, at least in this thread.

JR